Cross-platform store matching method and system based on multi-dimensional data fusion analysis

Abstract

The invention discloses a cross-platform store matching method and system based on multi-dimensional data fusion analysis, and relates to the technical field of data analysis, and the method comprises the steps: obtaining platform store multi-dimensional data from a plurality of data platforms; extracting store comprehensive features respectively corresponding to the multi-dimensional data of each platform store; performing weighted clustering on the comprehensive characteristics of each store to determine a plurality of corresponding clusters; in a weighted clustering process, calculating a corresponding membership degree according to the Euclidean distance between each store comprehensive feature and each clustering cluster; and for each store comprehensive feature, determining a matching cluster of the store comprehensive feature according to the cluster corresponding to the maximum membership degree, and determining the matching cluster as a cross-platform store matching result of the multi-dimensional data of the corresponding platform store. Therefore, through aggregation comprehensive analysis of the multi-dimensional data, the optimal matching result is output, and the consistency of O2O cross-platform matching results can be ensured.

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and in particular to a cross-platform store matching method and system based on multi-dimensional data fusion analysis. Background Art

[0002] With the booming development of the O2O (Online to Offline) industry, more and more companies rely on online and offline data linkage to improve user experience, optimize service processes and conduct precision marketing. Among them, store data matching has become a key link. Brands usually conduct business on multiple online platforms, and each platform has different ways of recording store names. In addition, the naming rules of the brands themselves are not unified, making it more difficult to match store data across platforms.

[0003] For example, a restaurant brand may use slightly different names on different platforms: "ABC Restaurant (Headquarters Store)" may be recorded as "ABC Catering Company Headquarters" on platform a, and may be named "ABC Main Store" on platform b. Although these differences are easy to identify from a user's perspective, it is a technical challenge for automated data systems to perform efficient and accurate matching. As a result, when O2O service providers integrate data, they often face the problem of inconsistency between the brand's offline stores and the online stores on various platforms, which seriously affects subsequent data analysis and service optimization.

[0004] In addition, a store’s geographic location data and brand characteristic data (such as the store’s monthly sales, ratings, etc.) may exist in different formats or standards on different platforms, further increasing the complexity of store matching.

[0005] In response to the above problems, the industry has not yet proposed a better technical solution. Summary of the invention

[0006] The present application provides a cross-platform store matching method, system, storage medium, computer program product and electronic device based on multi-dimensional data fusion analysis, which is used to at least solve the problem in the current related technology that cross-platform store data is difficult to effectively link due to inconsistent naming rules between different platforms.

[0007] In a first aspect, an embodiment of the present application provides a cross-platform store matching method based on multidimensional data fusion analysis, comprising: obtaining platform store multidimensional data from multiple data platforms respectively, wherein the platform store multidimensional data includes platform store name, platform store geographic location and platform store active time; extracting store comprehensive characteristics corresponding to each platform store multidimensional data; performing weighted clustering on each of the store comprehensive characteristics to determine a corresponding plurality of clustering clusters; in the weighted clustering process, calculating a corresponding membership degree according to the Euclidean distance between each store comprehensive characteristic and each clustering cluster; for each of the store comprehensive characteristics, determining a matching clustering cluster of the store comprehensive characteristics according to the clustering cluster with the corresponding maximum membership degree, and determining the matching clustering cluster as the cross-platform store matching result of the corresponding platform store multidimensional data.

[0008] In the second aspect, an embodiment of the present application provides a cross-platform store matching system based on multi-dimensional data fusion analysis, including: a data acquisition unit, used to obtain platform store multi-dimensional data from multiple data platforms, wherein the platform store multi-dimensional data includes the platform store name, platform store geographic location and platform store active time; a feature extraction unit, used to extract the store comprehensive features corresponding to each platform store multi-dimensional data; a feature clustering unit, used to weighted cluster each of the store comprehensive features to determine a corresponding plurality of clustering clusters; in the process of weighted clustering, the corresponding membership is calculated according to the Euclidean distance between each store comprehensive feature and each clustering cluster; a cross-platform matching unit, used to determine, for each of the store comprehensive features, a matching clustering cluster of the store comprehensive features according to the clustering cluster with the corresponding maximum membership, and determine the matching clustering cluster as the cross-platform store matching result of the corresponding platform store multi-dimensional data.

[0009] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the steps of the cross-platform store matching method based on multidimensional data fusion analysis of any embodiment of the present application.

[0010] In a fourth aspect, an embodiment of the present application provides a storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the cross-platform store matching method based on multidimensional data fusion analysis of any embodiment of the present application are implemented.

[0011] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of a cross-platform store matching method based on multidimensional data fusion analysis of any embodiment of the present application.

[0012] The cross-platform store matching method based on multi-dimensional data fusion analysis provided by this application can produce at least the following technical effects:

[0013] (1) Through the multi-dimensional data obtained from multiple platforms, including platform store names, geographic locations, and active time, a more complete store comprehensive feature is constructed, avoiding the errors that may be caused by single-dimensional data. For example, even if there are large differences in store names on different platforms (such as "ABC Catering Company Head Office" and "ABC Main Store"), they can be identified as actually corresponding to the same store through other dimensional information such as geographic location and active time. Therefore, through the comprehensive analysis of multi-dimensional data, the matching difficulties caused by inconsistent naming rules are solved, and the accuracy of cross-platform store matching is effectively improved.

[0014] (2) The inconsistency in recording store names on various platforms is one of the main challenges of store matching. Using a weighted clustering algorithm, the degree of membership between store comprehensive features and each cluster is calculated using Euclidean distance. When a store's comprehensive features have a certain degree of similarity with multiple clusters, the system will make a final matching judgment based on the principle of maximizing membership. Even if different platforms use different naming rules, stores with different names but the same essence can be effectively identified, significantly reducing the risk of mismatching.

[0015] (3) Through the clustering method of weighted clustering, it is allowed to adjust the weights of data of different dimensions according to the actual situation. For example, in areas where some stores are geographically dense, the feature influence weights of "store name" and "active time" can be increased to reduce the influence of geographical location. In the case of large differences in names, the feature influence weights of "geographic location" and "active time" can be enhanced. Therefore, through a flexible weight adjustment mechanism, the matching algorithm has dynamic adaptability, and can make intelligent adjustments according to the feature differences in different application scenarios, further improving the accuracy of cross-platform store matching results.

[0016] Through this technical solution, the optimal matching results are output through the aggregation and comprehensive analysis of multi-dimensional data, which can ensure the consistency of cross-platform matching results, reduce redundancy problems in cross-platform data analysis and processing, and effectively optimize the subsequent data analysis and service processes of O2O service providers. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flowchart of an example of a cross-platform store matching method based on multi-dimensional data fusion analysis according to an embodiment of the present application is shown;

[0019] Figure 2 Shown according to Figure 1 An exemplary operation flow chart of step S130 in FIG.

[0020] Figure 3 An operation flow chart of an example of adaptively adjusting the fuzzy index according to an embodiment of the present application is shown;

[0021] Figure 4 An operation flow chart of an example of determining matching clusters of comprehensive store features according to an embodiment of the present application is shown;

[0022] Figure 5 A structural block diagram of an example of a cross-platform store matching system based on multi-dimensional data fusion analysis according to an embodiment of the present application is shown;

[0023] Figure 6 It is a schematic structural diagram of an embodiment of an electronic device of the present application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0025] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0026] Figure 1 A flowchart of an example of a cross-platform store matching method based on multi-dimensional data fusion analysis according to an embodiment of the present application is shown.

[0027] Regarding the executor of the method of the embodiment of the present application, it can be any controller or processor with computing or processing capabilities. By combining multidimensional data, weighted clustering and membership calculation, it not only significantly improves the matching accuracy and robustness, but also has high adaptability and scalability. It can be widely used in various data integration and analysis scenarios in the O2O industry, ultimately promoting the improvement of user experience and the optimization of business processes.

[0028] In some examples, it may be integrated and configured in an electronic device or terminal by means of software, hardware, or a combination of software and hardware, and the type of the terminal or electronic device may be diverse, such as a mobile phone, a tablet computer, or a desktop computer, etc.

[0029] like Figure 1 As shown, in step S110, multi-dimensional data of platform stores are obtained from multiple data platforms respectively.

[0030] Here, the platform business type of the data platform can be a variety of O2O service types, such as e-commerce platforms, food delivery services, etc., and can be connected to each platform through an authorized API or data interface to automatically obtain multi-dimensional data related to the platform stores. In addition, the platform store multi-dimensional data includes the platform store name, platform store geographic location and platform store active time. The store name can be the display name of the store on each platform, taking into account the differences in naming rules that may exist on different platforms. The store's geographic location can be represented by latitude and longitude information or address information, and the geographic location information may use different coordinate systems (such as WGS84, GCJ-02, etc.). Store active time refers to the active period of the store on a certain platform, which may include business hours, rest time, and specific activity time, etc., which reflect the store's activity.

[0031] In some implementations, the obtained platform store multi-dimensional data can be uniformly formatted to ensure that the data on different platforms are structurally consistent. Specifically, the format of information such as store name, address, and time may vary from platform to platform, so data standardization is required to facilitate subsequent unified analysis and processing.

[0032] In step S120, the comprehensive store features corresponding to the multi-dimensional data of each platform store are extracted.

[0033] Here, for different types of data parameters in the platform store multidimensional data, corresponding feature extraction methods can be used respectively. Specifically, for the text features of store names, on the one hand, character similarity technology (e.g., Jaccard similarity) can be used to perform feature processing of store names. On the other hand, more intelligent semantic embedding technology (such as BERT or Word2Vec) can be used to convert each store name into a vector. In this way, not only the literal similarity in the store name can be captured, but also the semantic similarity can be captured. For example, "ABC Restaurant" and "ABC Food City" can be identified as similar store names through semantic embedding technology. For the extraction method of geographic location features, it can encode the geographic location, and can also enhance the recognition of location correlation through complex hierarchical geographic feature extraction, such as neighboring area division based on Haversine formula and fusion encoding of regional background information. For the extraction method of active time, it can integrate transaction behavior for feature extraction, such as extracting transaction active cycle, business hours, etc., to help effectively distinguish the transaction patterns of stores of different business types. Furthermore, in order to achieve more accurate store matching, the name text features, geographic features and time features corresponding to the platform stores can be integrated to obtain the corresponding comprehensive store features.

[0034] In step S130, weighted clustering is performed on the comprehensive features of each store to determine a corresponding plurality of clusters.

[0035] Here, in the process of weighted clustering, the corresponding membership is calculated based on the Euclidean distance between the comprehensive features of each store and each cluster. In weighted clustering, the comprehensive features of each store are weighted according to the weights of different features. The weight setting can be flexibly adjusted according to the importance of the features. For example, when the store names are very different, the weight of the geographical location and active time features can be increased. In addition, when the locations are dense but the names are relatively similar, the weight of the name text feature can be appropriately increased.

[0036] It should be noted that weighted clustering (or weighted fuzzy clustering) algorithms, such as WFCM (Weighted Fuzzy C-Means, weighted fuzzy C-means clustering) algorithms, are different from ordinary clustering algorithms (for example, K-means algorithm). Specifically, in ordinary clustering algorithms, the "hard partitioning" method is implemented, that is, each store data will eventually correspond to a cluster cluster, while in the WFCM algorithm, the concept of fuzzy membership is introduced, allowing a platform store data to belong to multiple cluster clusters at the same time, and each cluster has a membership to indicate its similarity strength with this cluster, and outputs multiple matching candidates according to different similarity values, which is particularly important for stores that are similar but not exactly the same. For example, the names and geographical locations of two stores are highly similar, but the business hours are different. In this case, ordinary clustering may completely ignore another store due to differences in a certain dimension, while weighted fuzzy clustering can assign different membership values ​​to the two stores, indicating that they may be fuzzy matches. In addition, the flexible weighting mechanism enables weighted fuzzy clustering to better handle multi-dimensional feature matching and improve the accuracy of overall data matching.

[0037] In step S140, for each store comprehensive feature, a matching clustering cluster of the store comprehensive feature is determined according to the clustering cluster with the corresponding maximum membership, and the matching clustering cluster is determined as the cross-platform store matching result of the corresponding platform store multi-dimensional data.

[0038] In some embodiments, for each platform store data, based on the degree of membership between it and each clustering cluster, the clustering cluster with the maximum degree of membership is selected as the affiliation of the store. In this way, if the maximum degree of membership corresponding to the data points on different platforms points to the same clustering cluster, it means that these data points represent the same actual store. As a further optimization, if the degree of membership corresponding to the store data is too fuzzy, the judgment can be assisted by adopting a threshold strategy or further adjusting the feature weights, and finally the data matching of the same actual store on different platforms can be achieved. Finally, the matching results can be output in the form of store names and other relevant features for subsequent data integration and analysis.

[0039] Through the embodiments of the present application, firstly, a rich store feature description is constructed by acquiring and extracting multi-dimensional data; secondly, based on the weighted clustering algorithm, information of different dimensions is comprehensively processed to ensure the accuracy of matching; finally, accurate matching of stores is achieved through membership calculation, and the final output matching result can greatly improve the efficiency and accuracy of store data integration in the O2O industry, and effectively realize accurate matching of store data across platforms.

[0040] Regarding the implementation details of step S120 above, in some implementations, the BERT embedding model is used to convert the store name into a vector so as to capture the semantic similarity of the store name, rather than just the character similarity, so that even if the store names on different platforms have differences in format or words, the system can perform more accurate matching through semantic similarity. In addition, for each store, its geographic location feature vector is a direct representation of its longitude and latitude values ​​to form a corresponding location feature. In addition, the time feature includes the store's business hours and active time period, so that the time feature of each store can be represented as a time series vector. Furthermore, by fusing the features of the three dimensions of name text features, geographic location features, and time features into a comprehensive feature vector, and using a weighted fusion approach to process the importance of each feature, the multi-dimensional feature fusion and matching can avoid the errors caused by a single dimension, and provide a more comprehensive and accurate matching result.

[0041] Specifically, the comprehensive store features corresponding to the multi-dimensional data of platform stores are extracted in the following ways:

[0042] v text (i)=BERT(G i ), formula (1)

[0043] v geo (i)=(lat i ,lon i ), formula (2)

[0044] v time (i)=(T i1 ,T i2 ,…,T in ), formula (3)

[0045] f(i)=α·v text (i)+β·v geo (i)+γ·v time (i), formula (4)

[0046]

[0047] Where f(i) represents the comprehensive features of the store i in the platform store multidimensional data, α, β and γ represent the weights of text features, geographic location features and time features respectively; v text (i) v geo (i) and v time (i) respectively represent the name text feature vector, geographic location feature vector and time feature vector of the platform store multidimensional data i; and They represent the variance of text features, geographic location features, and time features determined based on the feature distribution of multidimensional data of all platform stores; BERT represents the BERT model, G i Indicates the platform store name in i, lat i and lon i Respectively represent the longitude and latitude in i, T in Indicates the active state indicated by i in the nth time period.

[0048] In the embodiment of the present application, the degree of discreteness of the corresponding feature in the feature space is measured by variance. For example, the text feature variance is used to reflect the degree of discreteness in the semantic space. The larger the variance, the more dispersed the distribution of the store names and the greater the semantic difference. In addition, the variance of the geographic location feature can indicate the degree of discreteness of the store's geographic location. If the variance is large, it means that the distribution of the stores is more dispersed; if the variance is small, it means that the stores are more concentrated. In addition, the variance of the time feature reflects the degree of discreteness of the store's active time. The larger the variance, the greater the variation in the business hours or active time of different stores, and the stronger the discreteness.

[0049] Therefore, the larger the variance of a feature, the more dispersed the feature is in the entire store data set, which also means that the importance of the feature in the matching process can be reduced. In this way, by dynamically adjusting the weight, the matching algorithm can better adapt to the specific distribution of the data.

[0050] Through the embodiment of the present application, by adopting an adaptive weight mechanism, the weights of text, geographic and time features in matching are dynamically adjusted according to their variance. When a feature has a high degree of dispersion in the data set (i.e., a large variance), the system automatically reduces the weight of the feature in the comprehensive features, and implements a weight adaptive adjustment mechanism, which enables the system to automatically adjust the matching strategy in different scenarios. For example, in the case of large differences in the format of store names, the text features may not be reliable enough. At this time, the system automatically increases the weight of geographic location or time features, and reduces the influence of text features, thereby improving the adaptability of the matching algorithm and maintaining a high matching accuracy in different data set scenarios.

[0051] Figure 2 Shown according to Figure 1 An example operation flow chart of step S130 in FIG.

[0052] like Figure 2 As shown in the figure, by using the WFCM algorithm to perform fuzzy clustering on the features, each store is allowed to belong to multiple clusters at the same time, and the degree of membership to the cluster is assigned to each store data.

[0053] Specifically, in step S210, a membership matrix and a plurality of cluster centers are initialized.

[0054] Specifically, set the total number of clusters C and initialize each cluster center c k , where k = 1, 2, ..., C, initialize the membership matrix U and satisfy the following formula:

[0055]

[0056] In the formula, u ik It is the membership degree of the platform store multidimensional data i to cluster k, indicating the similarity degree of i belonging to cluster k.

[0057] In step S220, the Euclidean distance between the platform store multidimensional data and each cluster center is calculated to update the corresponding membership degree.

[0058]

[0059] Where m represents the fuzzy index, which is used to control the fuzziness of the membership, m>1; ||f(i)-c k || represents i and cluster center c k Euclidean distance; j represents the index of the cluster, ranging from j=1,2,...,C.

[0060] Here, due to the use of fuzzy membership u ik , stores no longer strictly belong to a certain cluster, but are associated with multiple cluster centers according to the degree of membership. This means that when the store name or other features are vague or uncertain, the system can provide multiple matching possibilities, increase the flexibility and robustness of the matching, and reduce the mismatch rate. For example, if the store names on two platforms are not exactly the same but have certain similarities, the cluster with a higher degree of membership can still reflect the potential matching relationship.

[0061] In step S230, each cluster center is updated according to the membership matrix of the current round.

[0062]

[0063] In the formula, is the fuzzy weight of the membership degree, N is the total number of multidimensional data of platform stores; c k It is the eigenvector of the center point of cluster k, indicating the center of gravity of the cluster; Represents the fuzzy weight of the membership of i to cluster k.

[0064] Here, based on the membership u ik To calculate the cluster center c kBy continuously iteratively calculating the Euclidean distance from stores to each cluster center and updating the membership degree and the location of the cluster center, the system can adaptively optimize the clustering results. k By updating the system, the boundaries of the clusters can be dynamically adjusted to ensure that the stores match the most appropriate clusters. By adjusting the weights of the cluster centers, clusters can be prevented from being dominated by abnormal data or outliers, thereby improving the stability of clustering. In addition, the adaptive adjustment of feature weights in the weighted fuzzy clustering method can reduce the occurrence of local optimal problems, ensuring that each iteration is reducing the Euclidean distance between the store and the optimal matching cluster, making the final clustering result closer to the global optimal solution.

[0065] In step S240, the updating steps of the membership degree and cluster center of the multidimensional data of each platform store are iteratively performed until the change in the cluster center is lower than a preset threshold.

[0066] Finally, through the above iterative convergence results, each store has a corresponding membership of one or more clusters. In this way, it is possible to extract comprehensive features of store data from cross-data platforms and cluster these stores through weighted fuzzy clustering. Through the adaptive weight adjustment mechanism, the importance of text, geographic and time features can be dynamically adjusted, and the ambiguity in the data can be handled, making the matching more accurate and flexible.

[0067] It should be noted that in fuzzy clustering, the fuzzy index m controls the degree to which a store belongs to multiple clusters, and by setting different m, the fuzziness of clustering can be effectively adjusted. For example, when m is large, the system can provide more candidate clusters for store matching, so that multiple matching possibilities can be referenced in uncertain situations, which is suitable for low-certainty scenarios and can provide multiple candidates. In addition, when m is small, the system is more inclined to hard matching, so that each store can be more clearly attributed to a single cluster, which can improve matching efficiency and reduce the consumption of computing resources in high-certainty scenarios.

[0068] In one example of the embodiment of the present application, the fuzzy index m can be customized by the user according to the distribution of the data to be analyzed. In another example of the embodiment of the present application, the fuzzy index m can automatically learn and identify the current clustering state and perform adaptive adjustment according to the clustering state.

[0069] Figure 3 An operational flow chart of an example of adaptively adjusting the blur index according to an embodiment of the present application is shown.

[0070] like Figure 3As shown, in step S310, the intra-cluster compactness is calculated based on the average value of the Euclidean distance between all store data and the center of the cluster to which they belong.

[0071]

[0072] In the formula, c k(i) represents the center point of the cluster k(i) to which the platform store multidimensional data i belongs, ||f(i)-c k(i) || represents the Euclidean distance between i and the center of the cluster to which it belongs; S intra It indicates the intra-cluster compactness. The smaller the intra-cluster compactness is, the stronger the similarity between the stores in the cluster is, and the better the clustering effect is.

[0073] Here, the intra-cluster closeness can be used to measure the similarity of the comprehensive characteristics of stores in the same cluster. When the intra-cluster closeness is poor (stores in the same cluster have low similarity), the system will increase the fuzzy index m so that store data is associated with multiple clusters to cope with complex matching situations and avoid making a single clustering decision too early.

[0074] In step S320 , the inter-cluster separation degree used to measure the differences between different clusters is calculated.

[0075]

[0076] In the formula, c k and c l Respectively represent the feature vectors of the center points of the kth cluster and the lth cluster, ||c k -c l || represents the Euclidean distance between two cluster centers; S inter It represents the separation between clusters. The larger the separation between clusters, the more obvious the differences between different clusters and the better the clustering effect.

[0077] Here, the difference between different clusters can be measured by the inter-cluster separation. When the inter-cluster separation is small (that is, the difference between clusters is not large), the fuzzy index will be increased accordingly to increase the fuzziness and reduce the occurrence of mismatches.

[0078] In step S330, the corresponding fuzzy index m is calculated according to the intra-cluster compactness and inter-cluster separation:

[0079]

[0080] Where λ is the adjustment factor, which is used to control the adjustment range of the fuzzy index.

[0081] Here, based on the ratio between the intra-cluster compactness and the inter-cluster separation, the quality of the current clustering can be evaluated. If the intra-cluster compactness is high (stores are more dispersed within the cluster) and the inter-cluster separation is small (there is no significant difference between different clusters), the clustering effect is poor, and the fuzzy index m needs to be increased to enhance the fuzziness. On the contrary, if the intra-cluster compactness is small and the inter-cluster separation is large, the clustering effect is good, and the fuzzy index can be appropriately reduced to make the system more deterministic. Therefore, by combining the intra-cluster compactness and the inter-cluster separation, the fuzzy index can be adaptively adjusted according to the feature clustering situation in each iteration to ensure that the differences in each dimension are effectively processed, so as to achieve a better balance of the influence of different dimensional features in the clustering process in the scenario of multi-dimensional feature fusion.

[0082] Through the embodiments of the present application, the fuzzy index m is automatically adjusted according to the actual distribution of the data, so that the system can deal with store data distribution with large similarity distinction (deterministic clustering) and store data distribution with strong ambiguity (flexible clustering). When the ratio of intra-cluster tightness to inter-cluster separation is large, the system automatically increases the fuzziness to avoid mismatching due to uneven data distribution. Dynamic adjustment of the fuzzy index m can help the system handle complex store matching scenarios, especially when there are inconsistencies or ambiguities in store names, geographic locations, and time features on multiple platforms, the system can still match stably. By dynamically adjusting m, the system can reduce ambiguity when the clustering effect is good, thereby avoiding clustering uncertainty caused by excessive ambiguity, and at the same time enhance ambiguity when the clustering effect is poor, avoiding clustering deviation caused by overfitting. Therefore, dynamically adjusting the fuzzy index based on data distribution can more effectively avoid local optimal problems.

[0083] Figure 4 An operational flowchart of an example of determining matching clusters of comprehensive store features according to an embodiment of the present application is shown.

[0084] It should be noted that in the actual cross-platform store data matching process, the multi-dimensional data of the store may have certain ambiguity or uncertainty. In order to better quantify such uncertainty and further improve the reliability of the matching results, a confidence threshold is introduced in this embodiment to optimize the fuzzy matching.

[0085] like Figure 4 As shown, in step S410, the matching confidence between the platform store multidimensional data i and the clustering cluster with the corresponding maximum membership degree is calculated.

[0086]

[0087] k * = argmax k (u ik ), formula (13)

[0088] In the formula, k * represents the cluster with the largest membership, k′ represents the cluster with the second largest membership, and δ i Represents i and k * The matching confidence between .

[0089] Here, using δ i To measure the matching confidence of each store, it can effectively distinguish those uncertain matching results. Compared with directly selecting the matching cluster based on the maximum membership, comparing it with the second largest membership, if the difference is small, it means that the store has a high correlation with multiple clusters at the same time, and the matching result is uncertain, which can help the system identify those matches with potential ambiguity.

[0090] In step S420, the matching confidence is compared with a preset confidence threshold τ.

[0091] Here, in order to ensure the accuracy of the matching results, a confidence threshold τ is pre-set in the system. i When the confidence level is greater than τ, the system considers the matching result to be credible; when the confidence level δ i When it is less than or close to τ, the system considers that the match is uncertain.

[0092] In step S421, when the matching confidence exceeds the confidence threshold, that is, in δ i >τ, the cluster k corresponding to the maximum membership * Determine the matching cluster corresponding to the multidimensional data i of the platform store.

[0093] In step S423, when the matching confidence does not exceed the confidence threshold, that is, in δ i When ≤τ, the candidate clustering cluster set is determined based on the preset number of clusters ranked top according to the corresponding membership degree.

[0094] By comparing the match confidence with the confidence threshold τ, the system can effectively handle these fuzzy matches. For matches with lower confidence, the system can mark these match results as "fuzzy matches" and recommend further manual confirmation or automated optimization processing (such as re-adjusting weights or recommending multiple match candidates), helping the system cope with complex data distribution scenarios in actual business and making the matching results more reliable.

[0095] In this way, when the confidence level is low, the system will not force the selection of the cluster with the largest membership, but will provide multiple possible matching clusters for the store, ensuring that the final decision is more flexible and accurate.

[0096] In step S430, corresponding interactive controls are generated for each candidate cluster in the candidate cluster set.

[0097] Here, users can manually select the most appropriate match from the candidate clusters, which enhances the user's control and participation in the system. Especially in complex or abnormal data scenarios, the system can allow users to participate in decision-making by generating candidate matching results, which can reduce the errors caused by automatic matching of the system.

[0098] In step S440, when a selection operation corresponding to a target interactive control is detected, the candidate cluster corresponding to the target interactive control is determined as the matching cluster corresponding to i.

[0099] It should be noted that when the data is complex or ambiguous, directly selecting the cluster with the maximum membership may not be the optimal solution. By introducing a set of candidate clusters, the system can display multiple clusters with higher membership. At this time, after marking the fuzzy matches with lower confidence, the user can make a selection based on multiple recommended candidate matching results through control interaction, so that the system can further adjust the cluster weights and matching rules based on user feedback.

[0100] For users or system administrators, using δ i Calculating the match confidence and setting the threshold helps users to more clearly understand the credibility of the match results. Especially in the case of low confidence, the system can clearly inform users of the uncertainty of the match and allow further review or confirmation, which enhances the controllability of the system matching process and the user's trust in the system decision.

[0101] In addition, for large-scale cross-platform data matching scenarios, by using the matching confidence δ i It can help the system quickly screen out high-confidence matching results and reduce unnecessary complex matching operations. When processing large-scale cross-platform store data matching, the system can first process those matches with higher confidence and quickly output some matching results; for those store data matches with lower confidence, there may be more complex confusion scenarios. At this time, after the system generates multiple candidate clusters, users can further confirm the matching results through interactive selection, reducing the risk of mismatching that may occur during automatic matching of the system. It can adapt to various complex cross-platform store data linkage scenarios, making the system more flexible.

[0102] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of actions combined, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0103] Figure 5 A structural block diagram of an example of a cross-platform store matching system based on multi-dimensional data fusion analysis according to an embodiment of the present application is shown.

[0104] like Figure 5 As shown, the cross-platform store matching system 500 based on multi-dimensional data fusion analysis includes a data acquisition unit 510, a feature extraction unit 520, a feature clustering unit 530 and a cross-platform matching unit 540.

[0105] The data acquisition unit 510 is used to respectively acquire platform store multi-dimensional data from multiple data platforms, where the platform store multi-dimensional data includes platform store name, platform store geographic location and platform store active time.

[0106] The feature extraction unit 520 is used to extract the comprehensive store features corresponding to the multi-dimensional data of each platform store.

[0107] The feature clustering unit 530 is used to perform weighted clustering on the comprehensive features of each store to determine a corresponding plurality of clusters; during the weighted clustering process, the corresponding membership degree is calculated based on the Euclidean distance between each comprehensive feature of the store and each cluster.

[0108] The cross-platform matching unit 540 is used to determine the matching clustering cluster of the store comprehensive characteristics according to the clustering cluster with the corresponding maximum membership for each of the store comprehensive characteristics, and determine the matching clustering cluster as the cross-platform store matching result of the corresponding platform store multi-dimensional data.

[0109] In some embodiments, an embodiment of the present application provides a non-volatile computer-readable storage medium, which stores one or more programs including execution instructions, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute any of the steps of the above-mentioned cross-platform store matching method based on multi-dimensional data fusion analysis in the present application.

[0110] In some embodiments, the embodiments of the present application also provide a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes any step of the above-mentioned cross-platform store matching method based on multidimensional data fusion analysis.

[0111] In some embodiments, an embodiment of the present application also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of a cross-platform store matching method based on multi-dimensional data fusion analysis.

[0112] Figure 6 is a schematic diagram of the hardware structure of an electronic device for executing a cross-platform store matching method based on multi-dimensional data fusion analysis provided by another embodiment of the present application, such as Figure 6 As shown, the device includes:

[0113] One or more processors 610 and memory 620, Figure 6 A processor 610 is taken as an example.

[0114] The device for executing the cross-platform store matching method based on multi-dimensional data fusion analysis may also include: an input device 630 and an output device 640 .

[0115] The processor 610, the memory 620, the input device 630 and the output device 640 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0116] The memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the cross-platform store matching method based on multi-dimensional data fusion analysis in the embodiment of the present application. The processor 610 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 620, that is, realizing the cross-platform store matching method based on multi-dimensional data fusion analysis in the above method embodiment.

[0117] The memory 620 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 620 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 620 may optionally include a memory remotely arranged relative to the processor 610, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0118] The input device 630 can receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 640 can include a display device such as a display screen.

[0119] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, the cross-platform store matching method based on multi-dimensional data fusion analysis in any of the above method embodiments is executed.

[0120] The above-mentioned product can execute the method provided in the embodiment of the present application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of the present application.

[0121] The electronic devices of the embodiments of the present application exist in various forms, including but not limited to:

[0122] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0123] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0124] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0125] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0126] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0127] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cross-platform store matching method based on multi-dimensional data fusion analysis, comprising: Acquire multi-dimensional data of platform stores from multiple data platforms respectively, wherein the multi-dimensional data of platform stores includes the name of the platform stores, the geographical location of the platform stores, and the active time of the platform stores; Extract the comprehensive store characteristics corresponding to the multi-dimensional data of stores on each platform; Performing weighted clustering on the comprehensive characteristics of each store to determine a corresponding plurality of clusters; in the weighted clustering process, calculating the corresponding membership degree according to the Euclidean distance between each comprehensive characteristic of the store and each cluster; For each of the comprehensive characteristics of the store, a matching clustering cluster of the comprehensive characteristics of the store is determined according to the clustering cluster with the corresponding maximum membership degree, and the matching clustering cluster is determined as the cross-platform store matching result of the multi-dimensional data of the corresponding platform store.

2. The method according to claim 1, wherein: The comprehensive store features corresponding to the platform store multi-dimensional data are extracted in the following ways: v text (i)=BERT(G i ), v geo (i)=(lat i ,lon i ), v time (i)=(T i1 ,T i2 ,...,T in ), f(i)=α·v text (i)+β·v geo (i)+γ·v time (i), Where f(i) represents the comprehensive features of the store i in the platform store multidimensional data, α, β and γ represent the weights of text features, geographic location features and time features respectively; v text (i) v geo (i) and v time (i) respectively represent the name text feature vector, geographic location feature vector and time feature vector of the platform store multidimensional data i; and They represent the variance of text features, geographic location features, and time features determined based on the feature distribution of multidimensional data of all platform stores; BERT represents the BERT model, G i Indicates the platform store name in i, lat i and lon i Respectively represent the longitude and latitude in i, T in Indicates the active state indicated by i in the nth time period.

3. The method according to claim 1, wherein: The weighted clustering of the comprehensive features of each store to determine a corresponding plurality of clusters includes: Set the total number of clusters C and initialize each cluster center c k , where k = 1, 2, ..., C, initialize the membership matrix U and satisfy the following formula: In the formula, u ik is the membership degree of the platform store multidimensional data i to cluster k, indicating the similarity degree of i to cluster k; Compute the multidimensional data of the platform store i and each cluster center c k The Euclidean distance of to update the corresponding membership: Where m represents the fuzzy index, which is used to control the fuzziness of the membership, m>1; ||f(i)-c k || represents the Euclidean distance between i and the cluster center ck; j represents the index of the cluster, ranging from j = 1, 2, ..., C; Update each cluster center according to the membership matrix of the current round: In the formula, is the fuzzy weight of the membership degree, N is the total number of multidimensional data of platform stores; c k It is the eigenvector of the center point of cluster k, indicating the center of gravity of the cluster; represents the fuzzy weight of the membership of i to cluster k; The updating steps of the membership degree and cluster center of the multidimensional data of each platform store are iteratively executed until the change of the cluster center is lower than the preset threshold.

4. The method according to claim 3, wherein: The fuzzy index m is adaptively determined by: The intra-cluster density is calculated based on the average value of the Euclidean distance between the multidimensional data of all platform stores and the center of the cluster to which they belong: In the formula, c k(i) represents the center point of the cluster k(i) to which the platform store multidimensional data i belongs, ||f(i)-c k(i) || represents the Euclidean distance between i and the center of the cluster to which it belongs; S intra Indicates the compactness within the cluster; Calculate the inter-cluster separation, which is used to measure the differences between different clusters: In the formula, c k and c l Respectively represent the feature vectors of the center points of the kth cluster and the lth cluster, ||c k -c l || represents the Euclidean distance between two cluster centers; S inter represents the degree of separation between clusters; According to the intra-cluster compactness and inter-cluster separation, the corresponding fuzzy index m is calculated: Where λ is the adjustment factor, which is used to control the adjustment range of the fuzzy index.

5. The method according to claim 3, wherein: Matching clusters are determined in the following ways: Calculate the matching confidence between the platform store multidimensional data i and the corresponding cluster with the maximum membership degree: k * =argmax k (u ik ), In the formula, k * represents the cluster with the largest membership, k′ represents the cluster with the second largest membership, and δ i Represents i and k * The matching confidence between ; Comparing the matching confidence with a preset confidence threshold τ; When δ i >τ, k * Determine the matching cluster corresponding to i.

6. The method according to claim 5, wherein: After comparing the matching confidence with a preset confidence threshold τ, the method further includes: When δ i When ≤τ, the candidate clustering cluster set is determined according to the preset number of clusters ranked top according to the corresponding membership degree; For each candidate cluster in the candidate cluster set, generate a corresponding interactive control respectively; When a selection operation corresponding to a target interactive control is detected, the candidate cluster corresponding to the target interactive control is determined as the matching cluster corresponding to i.

7. A cross-platform store matching system based on multi-dimensional data fusion analysis, comprising: A data acquisition unit, used to acquire platform store multi-dimensional data from multiple data platforms, wherein the platform store multi-dimensional data includes platform store name, platform store geographic location and platform store active time; A feature extraction unit is used to extract comprehensive store features corresponding to the multi-dimensional data of stores on each platform; A feature clustering unit is used to perform weighted clustering on the comprehensive features of each store to determine a corresponding plurality of clusters; in the weighted clustering process, the corresponding membership degree is calculated according to the Euclidean distance between each store comprehensive feature and each cluster; The cross-platform matching unit is used to determine the matching clustering cluster of the store comprehensive characteristics according to the clustering cluster with the corresponding maximum membership degree for each of the store comprehensive characteristics, and determine the matching clustering cluster as the cross-platform store matching result of the multi-dimensional data of the corresponding platform store.

Images